Image Generation 벤치마크
Image Generation on CLEVR
FID-5k-training-steps 낮을수록 좋음
- 2021-03-01 — GANformer: FID-5k-training-steps 9.1679
- 2021-11-01 — Projected GAN: FID-5k-training-steps 0.89
| Rank | Model | FID-5k-training-steps | Paper | Code | Year |
|---|---|---|---|---|---|
| 1 | Projected GAN | 0.89 | Projected GANs Converge Faster | autonomousvision/projected_gan · dome272/ProjectedGAN-pytorch · tsubota-kouga/ProjectedGAN | 2021 |
| 2 | GANformer | 9.1679 | Generative Adversarial Transformers | dorarad/gansformer · lucidrains/transganformer | 2021 |
| 3 | StyleGAN2 | 16.0534 | Generative Adversarial Transformers | dorarad/gansformer · lucidrains/transganformer | 2021 |
| 4 | GAN | 25.0244 | Generative Adversarial Transformers | dorarad/gansformer · lucidrains/transganformer | 2021 |
| 5 | SAGAN | 26.0433 | Generative Adversarial Transformers | dorarad/gansformer · lucidrains/transganformer | 2021 |
| 6 | VQGAN | 32.6031 | Generative Adversarial Transformers | dorarad/gansformer · lucidrains/transganformer | 2021 |